Earlier this month, at 04:17 UTC, I watched a transaction from an OFAC-flagged Ethereum address sit in the mempool for 47 minutes. No relay advertised it to the builder network. No builder included it. The chain kept producing blocks — 145 of them — while that transaction waited. It never confirmed.
That is the quiet reality behind a fresh essay now circulating through Ethereum research circles. It asks a question the industry thought it had buried under a recovery-cycle bull narrative: who actually decides what gets onchain? The piece arrives without a technical proposal, without an EIP reference, without a clean answer. That is exactly why it matters. When someone deep in protocol theory reopens a wound without offering a bandage, they are not starting a conversation. They are flagging a structural fracture.
The answer today is not “the Ethereum protocol.” It is a small cluster of commercial block builders operating through the MEV-Boost pipeline. The essay is asking whether that arrangement still deserves its seat.
The State of Play
Let’s get the mechanics on the table. Ethereum separates “proposing a block” from “building a block.” That is proposer-builder separation, PBS by design. Validators — the entities that finalize state — stay lean. Builders compete to assemble the most profitable blocks, then hand them to validators for signatures. Market competition was supposed to keep the system efficient. Instead, it created a chokepoint. The logic was sound in 2020. The market that grew around it was never part of the whitepaper.
The data has been public since the Merge in September 2022. The overwhelming majority of Ethereum blocks flow through the MEV-Boost middleware, and within that system, a handful of builders and relayers control the overwhelming share of block construction. At times, a single relay has carried more than 70% of MEV-Boost blocks; the top few builders consistently assemble the majority of what reaches the chain. Build counts, relay statistics, inclusion rates — all open. Concentration this consistent is not an accident. It is the shape of an emergent industry with gatekeeping powers. For context, the top builders routinely produce more blocks than all other validators combined. That is not a failure of PBS. It is PBS working exactly as the market demanded.
The pattern in the logs is straightforward: most validators do not inspect payloads. They sign whatever the relay delivers. The separation of roles was supposed to reduce validator intelligence requirements. It also made validators functionally blind. The entity signing the canonical chain cannot tell you what transactions are inside the block. That is not a bug. It is the design.
Then came the precedent that never got a formal hearing. October 2022: the Treasury’s OFAC sanctioned Tornado Cash. Protocol-level Ethereum changed nothing. The middleware layer changed everything. Relayers refused payloads that contained transactions touching sanctioned addresses. Builders stopped touching them. The network kept building blocks. The chokepoint revealed itself without a single EIP.
Community designers responded with Inclusion Lists — a proposal allowing validators to force specific transactions into blocks regardless of builder preferences. Years later, it still is not deployed on mainnet. The protocol never answered the question it stumbled into. So the question festered.
That is the context for this essay. It asks “who decides?” more honestly than most. The answer is nobody, and everybody, and mostly the people who run relays.
Three Problems Masquerading as One
Now the forensic part. The essay merges three separate problems into one, and anyone framing it as a single question will get a single, oversimplified answer back.
First problem: block construction is a market, not a protocol output. When you submit a transaction, it lands in the mempool. Builders scan it, calculate expected value, and cherry-pick what matters. In my own monitoring work — from the Shanghai upgrade forward — I have logged the same pattern repeatedly. Builders are ruthless arbitrageurs. Unprofitable transactions wait. OFAC-flagged transactions wait longer. Some never get in. This is normally invisible to wallet users, but the relay logs do not lie. Censorship on Ethereum is not enforced by the consensus layer. It is enforced by a small cluster of off-chain entities the market happens to trust.
I built my own relay monitoring stack in 2023 because public dashboards lag. The monthly build-count tables do not show exclusion dynamics. They do not show which transactions were left out. Only the private order flow knows that. What I saw consistently was a two-tier market: high-value transactions get fast-tracked through specialized order-flow agreements; everything else waits for a builder with leftover block space. That is normal MEV behavior. But it means “neutral inclusion” was never the baseline. It was always a luxury of low regulatory pressure.
Second problem: the sanctions ghost. Current filtering behavior mirrors the Treasury’s SDN list far more than it reflects any protocol rule. Read that again: a US government agency is effectively setting inclusion policy for a network that markets itself as neutral settlement infrastructure. Nobody voted on it. The builders never published a governance proposal. It happened because the dominant relay infrastructure decided that being OFAC-compliant was better for business. That is not a conspiracy. It is an incentive structure. But it rewrites the meaning of “who decides.”
The most cited example remains the 47-minute mempool wait I logged this month. The transaction did not get rejected. It simply never happened. That is the signature of relay-level filtering: nothing is denied, so nothing can be flagged. Try proving censorship of a transaction that was never included. Try auditing a process with no receipt. That is where the debate gets stuck.
Third problem: the economics of curing it. Inclusion Lists would theoretically force builders to include transactions a validator flagged. Fine in principle. But make it a protocol mandate, and you hit the real wall — the staking economy. Validator income is a composite of consensus rewards, transaction fees, and MEV tips. Institutional stakers — Coinbase, Lido-affiliated node operators, Kraken — are regulated entities with compliance departments. Lido alone commands roughly a third of staked ETH. That is not a neutral observer; it is a protocol-level power center with legal exposure across multiple jurisdictions. A protocol obligation to include transactions OFAC has designated is a direct legal collision: a US-licensed company either provides a “material service” to a sanctioned entity, or violates the network’s rules. That is a lawyer’s dream and a validator’s nightmare.
The institutional response would not be clever. It would be withdrawal. Node operators resign. Validators consolidate into jurisdictions beyond US reach. The network ends up more concentrated, not less. The current ambiguity — the fact that no one formally decides — is exactly what allows big capital to keep participating. Formalize the decision, and everyone must pick a side. In that world, the “less centralized” network is the one where inclusion rules stay messy. The consensus layer has no variable for that nuance.
Fourth problem: the governance void. People keep writing “Ethereum should...” as if a steering wheel exists. It does not. A protocol-level answer would require an EIP, client-team implementation, validator signaling, and a coordinated network upgrade. That is multiple independent power centers, each able to say no. This essay poses its question to no one in particular because that is the actual answer: no one owns this decision. It gets made by default, inside the MEV-Boost pipeline, by actors whose incentives are mostly commercial. Anyone waiting for core developers to “solve” this misunderstands where the power sits.
And the default path — the one nobody is choosing — is drift. The essay will generate forum threads and a few conference panels, then dissolve back into the background. That is the most likely outcome. But drift is not neutrality. Every month that passes without an answer, the answer is being written by the builders, the relayers, and the sanction lists. The absence of a decision is a decision.
There is also a second-order consequence hiding in plain sight: Layer 2s. Every rollup settles to Ethereum L1 eventually. If L1 inclusion behavior turns formally censored — or formally uncensorable — every L2’s security assumptions shift. Rollups do not care whether the exclusion came from a builder preference or a law. They care whether their transactions settle reliably. Under the current ambiguity, that is survivable. Formalize the rule, and you push risk into every downstream protocol at once. The blast radius is larger than the essay acknowledges.
The Case for Keeping It Ambiguous
Here is the view nobody in the room wants to say out loud: full protocol-level censorship resistance would probably make Ethereum less safe, not more.
Walk the logic. Once the protocol mandates inclusion of all transactions, it stops being neutral. It becomes an affirmative refusal to cooperate with sanctions law. US regulators do not need to ban Ethereum to make it toxic for institutional capital. They just classify validators and nodes as providing “substantial services” — and the mandatory-inclusion requirement becomes the evidence. The conservative reading of OFAC’s 2022 guidance already treats the question of whether software infrastructure counts as a “material service” as open. A validator running mandatory-inclusion code would be the clearest test case the regulators could ask for. This is not paranoia. Europe’s MiCA is already pushing extraterritorial expectations around transaction processing, and the compliance treadmill keeps accelerating.
The darker blind spot: the anti-censorship camp obsesses over the builder chokepoint while ignoring that the chokepoint’s murkiness is protective. As long as “who decides” has no clean answer, governments cannot easily target a named decision-maker. The moment Ethereum formalizes an answer, it hands regulators a face, a mechanism, and a legal hook. The ambiguity is armor. That is cold. It is also why this debate keeps returning and never resolves.
The deeper trap: the essay assumes the visible mempool stays the battlefield. It will not. The industry is already moving toward encrypted mempools and private order flow. In that world, filtering becomes invisible by default — not because of OFAC compliance, but because no one outside the builder can see pending transactions at all. That kills the debate in the opposite direction: less censorship you can prove, not more censorship resistance you can guarantee.
And that matters more in a bull market, not less. Euphoria discounts tail risk. The last cycle taught everyone what happens when the tail arrives. This cycle is pricing the settlement layer as if its neutrality is settled law. It is not. It is a market outcome, enforced by relays, paid for with MEV tips. Every cycle, the same sequence plays out: regulators move, the community debates, builders adjust their filters, and the market moves on. Nothing about that sequence requires protocol change.
What to Watch
Stop watching the price chart. Start watching the AllCoreDevs agenda and the Ethereum Research forum. If an EIP referencing Inclusion Lists or a variant picks up validator support, the structural risk becomes a market event. If Coinbase or Lido publishes a position on mandatory inclusion, that is your trigger. Until then, this essay is less a forecast than a mirror. The answer to “who decides” was decided long ago — by relay logs and builder incentives. The only open question is whether anyone bothered to look. And if you run a validator yourself: stop signing blind. Start reading the payloads you approve.
Timestamped. Traced. Published. The chain doesn’t lie. Incentives do. Speed is a weapon. Data is the trigger.
